You add up the mass of raw materials your facility consumes each day, look at the products it’s made, and the numbers just don’t balance—until you look at skids of rework and a dumpster topped off with the day’s malformed products and/or damaged packages. What went wrong, and how do you get a handle on it?

Article Index:

• It starts with the ingredients
• Beyond the low-hanging fruit
• Software solutions
• Methods and procedures to find yield problems
• Prepare for improving yield
• Is historical data needed?

There are many places in a food plant where yield can be improved and, most likely, quality too. Some are more obvious than others. Sometimes, a multitude of factors comes together and stacks up against a day’s production. Issues may be found, for example, in controlling the ingredients being combined into the recipe, filling and packaging—not to mention controlling temperatures in critical applications. Fill weights, quality and food safety can also affect yield. But finding the root causes of these yield problems takes an investment in measurement and control solutions, which have been shown to pay for themselves rather quickly.

When it comes to finding solutions to increase yield, it may seem all the “low-hanging fruit” has already been picked.  But for many processors, some of that low-hanging fruit has been missed, and it can easily be found by making some measurements. The old adage, “You can’t control what you don’t measure,” still applies.

“There are still gains to be realized in managing the risky aspects of operations—from line startup to changeovers to new product introduction,” says Eric Lemaire, Schneider Electric Global Solutions director, food and beverage strategy. “Directing the steps and activities, monitoring for problems, detecting leading indicators or an actual event and notifying and orchestrating a response can become a competitive advantage to those who invest in a sense-and-respond solution.”

According to Ron Pozarski, senior marketing manager for Rosemount Flow Products, Emerson Process Management, process optimization and maintaining an optimized operation over the life of the process can significantly improve and maintain yield, increase uptime and reduce energy usage. But processors often find themselves trying to reduce downtime while coping with reduced effective capacity due to both internal and external regulatory compliance requirements.

“To improve yield, the focus of food manufacturers is usually on optimizing the areas of an operation that cause a production run to be outside the statistical control limits,” says Arvind Rao, Rockwell Automation NA food and beverage industry commercial manager. “A less common focus is refining the control limits, which allows manufacturers to tighten their process in order to maximize yield.”

However, many food manufacturers’ systems may be unable to maintain process control with more refined constraints, adds Rao. “This lost opportunity to improve yield is why an advanced process control system is important for manufacturers that need to increase efficiency over a large variety of product lines to remain competitive.”

It starts with the ingredients

“The most common raw material loss areas in the food and beverage industry are related to net-weight fill and changeovers,” observes Katie Moore, GE Intelligent Platforms global industry manager. “Consequently, many manufacturers focus on these first.” Moore, who once served as a food plant manager, has seen these issues firsthand. The limitations around net-weight fill may be due to the precision (or lack thereof) of the equipment processing the products. For example, when a certain large snack food manufacturer employed tools to check fill, daily waste went down from 2,000 to 200 lbs., film was saved, and net-weight fill optimization resulted in six to seven figures of dollar savings per plant.

Moore offers another example: In baking, a processor may have a divider that doesn’t measure each dough ball precisely. In this case, modifications or improvements may need to be made to the divider to improve precision. But first, finished package weights would need be measured to determine the baseline “giveaway” loss.

When it comes to changeovers, not executing them effectively may result in more loss or “bad” product being made from one run to another. But, implementing continuous improve­­ment methodologies along with software technologies to measure this data automatically can begin to shine a light on opportunity areas, adds Moore.

With complex systems, software can be an issue in itself. “Developing software to automate material deliveries can quickly become complicated when programming a process control system,” says Randy Otto, vice president of ECS Solutions, a member of the Control System Integrators Association (CSIA). “Systems with valve clusters that allow ingredients to be sent to multiple vessels can complicate the software.” Common solutions to obtaining accurate deliveries include slowing the delivery of a material throughout the delivery cycle or toward the end of the delivery cycle (dribble flow or dribble speed). “Herein lies the issue,” continues Otto. “The end-user must sacrifice speed [performance] to obtain quality. Overall yield is a product of performance and quality, so reducing one to increase the other produces a minimal result. Yield is significantly improved when you can increase performance or quality while maintaining or increasing the other.”

“Performance issues are reflected back to yield and quality opportunities—and even more on high-speed production lines,” says Tom Braydich, senior consultant at Matrix Technologies, a member of CSIA. Some of the issues are improper line balancing to prevent surges and stoppages, ill-maintained equipment, product and material supplies (cans, bottles, cartons, labels, etc.).

Matrix Technologies can help with an overall equipment evaluation and modification of filling and packaging systems to ensure line coordination among OEM vendors. “When a line is designed so all the equipment operates as one coordinated unit [regardless of the OEMs involved], the throughput can increase dramatically,” adds Braydich. “Large improvements can be realized through a simple investment in automation for the overall line versus the individual machines.”

Other issues also can complicate raw ingredients and formulation. “Production yield for a given formulation varies depending on the raw ingredients being used [e.g., the quantity of denatured proteins in milk affecting yield in cheese making],” says John Tertin, director of manufacturing information systems, ESE Inc., a member of CSIA. “Making in-process adjustments to account for variance in the raw materials can recover yield that would otherwise be lost in processing.”

Another issue related to measuring ingredients is the mating of piping with the flow meters in a system. “When adding new flow meters to improve data collection, we check the piping design to improve the performance of the flow meters,” says Robin Cobb, project manager at Polytron, Inc. “We often find a piping design is not conducive to the best practices for measuring density set points.” Related to mass flow meters, Cobb suggests automating density set points to ensure standardization of how density is measured, taking it out of the operator’s hands and guaranteeing accurate, apples-to-apples comparisons.

Another common issue is the control of product loss at transfer points, according to Arun Madangarli, Polytron senior engineer. “In a beverage process, for example, loss typically occurs in two areas: the unit operations that are part of the process and during the transportation of product through the process line when it is transferred from one stage to the next. Over time, the loss at these points can be a significant amount.”

For a small operation, however, waste can be improved with relatively simple automation. For example, a bag filler can be automated with a weigh scale and a small PLC and HMI in the same housing, according to Don Pham, IDEC product manager. The IDEC combined HMI+PLC unit provides the automatic cycle and process control required for a bag filler and frees up the operator to tend to other plant operations. Variable fill rates and impeller speeds are controlled by the unit based on readings from the electronic weigh scale. The automation assures each bag will be filled to the correct level, without overfill.

Beyond the low-hanging fruit

Sometimes, figuring out what is causing a yield and/or quality problem can be difficult, but often, it is very process specific and can be tied in with ingredients. “If the product is baked, there may be issues around a consistent, thorough bake in the oven,” states GE’s Moore. “Another partially ambiguous loss area—or one that may be harder to pinpoint—is the actual effect of varying raw material specifications on finished products. For example, in baking, slight fluctuations in protein in the flour [negative or positive] affect yield. This can be hard to determine if the specifications aren’t tied to the lot digitally.”

Incorporating actual raw material properties into the production process saves time and ensures product quality. “For example, in an extraction process, the amount of water needed to reach a desired finished product total acidity [TA] is dependent on the TA of the raw ingredient,” says ESE’s Tertin. “You could add about 80 percent of the water, test to determine how much more is necessary and add the remainder. But, [if you incorporate the quantity and TA of the raw material into the process], you can add exactly the right amount of water to begin with and ensure consistent yield and save time in that portion of the process.”

Human factors, such as a lack of process knowledge, are additional considerations, according to Carl Harper, ICONICS project engineer. “Increasing set points to make production line or individual machines run more quickly to meet targets affects quality. But, without process knowledge, employees may believe, ‘If the line/equipment is running, we are in control, and all is well.’ Consequently, when it isn’t running properly, nobody knows why or how to fix it.”

In addition, when working with dense products like soup, bread and cookie dough, end-of-line checkweighers can provide a false sense of control over product final weight, says Matrix Technologies’ Braydich. “The fluctuation of weights may be the result of recipe errors due to a faulty flow meter, scales or worker behavior. Whatever the cause, an operator will repeatedly adjust the filler or depositor weight all day long. The consumer will get the correct weight, but the quality of the product will be the big question.”

Maintenance and control software also can have a positive or negative effect on yield. “Processors are becoming increasingly aware that reliability centered maintenance [RCM] can keep equipment in optimal operating condition and provide the highest availability [improving yield and meeting quality targets],” says Schneider’s Lemaire. “Technology developed and used in heavy industries is finding its way into food and beverage processing plants’ maintenance practices.” (For more on RCM and predictive maintenance, see “The future looks smarter,” FE, July 2015.)

Software solutions

Control software plays a big part in improving yield. However, more obscure yield and quality problems frequently can be masked or made unattainable by just-average controller software, according to ECS Solutions’ Otto. “Controller software is only as good as the engineer who developed it, and often, the developer is under critical constraints [e.g., time, budget or scope issues] that prevent him or her from developing the best solution. The controller software can become complicated, especially over time with multiple developers adding to it, creating greater challenges and barriers to make improvements to the system performance.”

Many processors turn to commercial off-the-shelf (COTS) software packages that log, analyze and report data in numerous ways to identify the greatest problems in a system. Otto says that while MES packages with these options promise great benefits, they are costly to purchase and implement. Plus, while they may identify areas of constraints, they seldom offer solutions. ECS has implemented numerous flexible solutions, replacing or enhancing controller software, and providing a means for end-users to optimize performance gradually. “Our customers have discovered low-hanging fruit they were unaware of, realizing performance increases of 20 to 40 percent, as well as improved quality and a 5 percent increase in availability,” concludes Otto.

Implementing a manufacturing intelligence solution enables long-term process improvements that can help make a plant more flexible and improve margins. “Tyson Foods was experiencing unacceptable variances in the weight of Hillshire Brands plastic-wrapped breakfast sausage, or ‘chubs,’ causing waste and rework,” recalls Rockwell’s Rao. “The company turned to FactoryTalk software to better understand exactly where the variance issues were being created. This [software] update changed Tyson’s available data from an ex post facto report on how much waste was produced to dynamic, real-time data on weight fluctuations along the line, the exact giveaways on finished product, machine downtime, OEE and meat temperature. The plant ultimately saw an increase in yield of 0.50 percent, which is more than half a million pounds of sausage annually.”

Methods and procedures to find yield problems

What methods or systems should processors employ to find and fix yield issues? “First and foremost, are the collection of key information [set points, temperatures, weights, flows, counts, etc.] in the process and linking it by time, schedule and batch,” says Braydich. “Simple SPC [statistical process control] methods like upper and lower control limits applied against the data will identify variation in blending and batching processes.” Braydich reports food processors realizing ROIs from 20 to 30 percent. Typically, yield improvements precipitate savings in other areas like OEE.

Using SPC helps operators monitor processes in real time and reveals process variations including out-of-control and trend issues, says ICONICS’ Harper. In addition, performing 100 percent inspections allows users to capture and store more data. SPC informs operators of process stability and capability by exposing key metrics like Cpk (process capability) and Ppk (process performance). Harper also suggests implementing a solution for Six Sigma, ISO 9000 and Total Quality Management (TQM) initiatives. In addition, using fault detection and diagnostics (FDD) technology helps identify faults in the process (e.g., an oven temperature greater than x° for five seconds). An FDD solution can also point out energy surges and detect how many times they occur.

A mass-balance reporting system that encompasses the entire production line (from bulk material intake to finished product storage) is another method that can track and quantify product loss, says Polytron’s Madangarli. “While a typical report measures yield in terms of units relevant to the corresponding process area [gallons, cases, bottles, etc.], mass-balance reports compute yield in units of mass [lbs. or kg]. This gives a much more accurate picture of the process yield, especially if a precise, real-time measurement of the product density is available.”

ESE’s Tertin recommends using advanced control techniques and inline instrumentation plus maximizing data integration between the ERP/MES and control system layers—either directly (in a non-isolated networked environment) or by the use of barcodes/RFID tags. SPC calculations are part of ESE’s Q5 analytical software package. One Midwest cheese processor achieved a complete payback on a Q5 analyzer in 150 days, based on QC consumables and lab hard-cost savings.

“For material deliveries within process systems, ECS uses statistical spill value tracking to meter every ingredient to every location,” explains Otto. “With this method, we can not only substantially increase accuracies in material deliveries, which improves quality, but also increase yield by calculating a very accurate spill value at full speed.” By employing this method, one food processor recently realized a 30 percent increase in performance, improved quality and a 5 percent greater equipment availability.

“Advanced process control [APC] and model predictive control [MPC] are effective in increasing process yield and reducing out-of-spec product at the beginning and end of a process run,” says Emerson’s Pozarski. “Precision measurements can decrease overfill without risking under-fill in packaging and filling applications.”

While APC or MPC applications aren’t for everybody, they can be very effective in certain applications. According to Pozarski, a major brewery reduced beer loss by 1 percent, saving over $70,000 per 1 million barrels brewed, by identifying the unexpected loss of yield in a process unit and correcting the root cause. In another case, a beverage processor eliminated a process bottleneck in its mix room, increasing available capacity by a factor of four,  by updating its traditional batch weigh system to one that simultaneously blends liquid components into the mix tank through precision flow meters.

“Many software packages that may provide incremental value have different capabilities including SPC and nonconformance management,” says GE’s Moore. “But, the fundamental priority for a manufacturer is to make sure the system is built on an open platform to be able to access the data and work with other systems for scalability and repeatability purposes. A standalone, disparate system may provide initial value, but over time, the cost to manage and maintain another system outweighs the benefits.”

Prepare for improving yield

Sensors are at the edge (or periphery) of holistic systems that measure plant operations, says Schneider’s Lemaire. They sense and measure the process and primarily detect when things go wrong.

At a minimum, control systems, as well as some automation and instrumentation, should be installed wherever possible, observes Moore. A processor must be able to measure product flow as it moves throughout the entire process. Software can then be added to aggregate and correlate that data and information.

In many applications, weigh scales can be upgraded or replaced to obtain this data, says Braydich. Plus, flow meters can both improve throughput and increase accuracy in batch applications. When materials can be added simultaneously, batch time can be reduced significantly while improving consistency between the batches.

According to Tertin, depending on the process area, additional hardware may include analytical instrumentation, 2D barcoding systems, Coriolis flow meters, proximity sensors and smart devices such as Ethernet-connected VFDs, weight controllers, load cells, X-ray systems and metal detectors.

Fortunately, new systems come with sensing devices, but sometimes, they require a little adjustment. For example, ECS’s Otto describes a recent optimization of a bourbon bottling line: “The optimization required software changes and moving existing bottle sensors around on the line to increase performance and quality. New and additional VFDs and sensors were recommended to help reduce impacts of the glass bottles and further optimize the performance.”

Integrating sensors with process control systems also improves a system’s performance. Otto notes that one CIP system delivered by ECS with integrated sensors and software helped the end-user reduce CIP time from 13 hours per week to four hours per week, improving the availability and yield of the process equipment.

Is historical data needed?

One final question: Does a processor need a repository of archived process data to install software systems that can monitor yield and performance? “Historical data will always be helpful in implementing new systems, although new systems for improving yield have been implemented without it,” answers Rockwell’s Rao. What makes historical data significant is its ability to provide context to the improvements made by a new system. For instance, historical data can set a good baseline for understanding how much value a project will add.

“Without ‘hard’ data, there will always be a certain amount of conjecture regarding the potential gain [of a new project],” says Tertin. “For some customers, this is acceptable, given their desire for short-term, incremental improvement; for others, it is not.”

Historical data can be used to compare improvements made by automating a particular work step. For instance, with a bagging machine, the data could include the amount of labor necessary to operate it manually, the required rework from operator errors and the reduced throughput due to waiting on required operator actions, according to IDEC’s Pham.

GE’s Moore suggests that while a processor does not need to have any historical data to begin implementing a system, reviewing the data should be the first step in looking at technology to aid yield improvement. A processor should ask three questions about collected data before embarking on this process: Am I collecting meaningful data? Do I have a strategy to store and access the data I collect? Have I prioritized the outcomes I desire?

For more information:

Eric Lemaire, Schneider Electric, 949-727-3200, eric-s.lemaire@schneider-electric.com, www.schneider-electric.com

Arvind Rao, Rockwell Automation, 440-646-3434, avrao@ra.rockwell.com, www.rockwellautomation.com

John Tertin, ESE Inc., 715-387-4778, tertinj@ese1.com, www.ese1.com

Katie Moore, GE Intelligent Platforms, 800-433-2682, katie.moore@ge.com, www.ge.com

Randy Otto, ECS Solutions, 812-479-5170, randy.otto@ecssolutions.com, www.ecssolutions.com

Robin Cobb, Polytron Inc., 855-794-7659, rcobb@polytron.com, www.polytron.com

Arun Madangarli, Polytron Inc., 855-794-7659, amadangarli@polytron.com, www.polytron.com

Tom Braydich, Matrix Technologies, 419-897-7200, ext. 306, tmbraydich@matrixti.com, www.matrixti.com

Don Pham, IDEC, 408-747-0550, donp@idec.com, www.idec.com

Carl Harper, ICONICS, 508-543-8600, carl.harper@iconics-uk.com, www.iconics.com

Ron Pozarski, Emerson Process Management, 952-828-3434, ron.pozarski@emerson.com, www.emersonprocess.com

Frank Tappen, DataNet Quality Systems, 248-447-0120, ftappen@winspc.com; www.winspc.com

Dave Boeldt, Bosch Rexroth, 1-800-REXROTH, dave.boeldt@boschrexroth-us.com; www.boschrexroth.com