Pistachio Roasting with Precision

With a trusted method for steam pasteurization and identification of a class problem in nut roasting, - we observed product inconsistency in the market, from over-roasted and burnt, to under-roasted and mushy. We set out to leverage our work in forced convection to deliver the Precision Roasting system. The Precision Roasting system is a dry-roasting continuous roaster that uses forced convection to eliminate cold spots and variability when dry-roasting various tree nuts, including almonds, pistachios, cashews, macadamias, pecans, hazelnuts, and peanuts.

Measuring and Modifying the Moisture and Color of Roasted Pistachios

Introduction

Laitram brings a scientific approach to process optimization that benefits the nut industry. Two methods were developed to obtain faster, more accurate, and repeatable ways to objectively measure the moisture and color of ground nuts. Moisture and color are two important indicators of quality and yield that are impacted by roaster design and operating conditions. These methods aided the effort to optimize the Precision Roaster's design and operation.

These methods were used to measure the moisture and color of market samples to investigate industry norms and variability (see Table 1). We conducted a matrix of roasting tests to explore the impacts on roasted moisture and color by the incoming raw material and roaster time and temperature settings.

This whitepaper explains why and how we developed these measurements and shares our market sampling and roaster test matrix results.

Measurement Improvement Opportunities

The industry typically uses "weigh-dry-weigh" scales to measure nut moisture. These types of devices also referred to as "moisture balances" or "evaporative sensors," are available from multiple suppliers, including Ohaus (Figure 1) and Mettler-Toledo. This type of device first weighs the nuts, then uses a heat source to dry them, and then weighs them again at the end to determine the moisture loss. The "weigh-dry-weigh" approach is very time-consuming and sensitive to many factors, including the measurement time (faster is less accurate), the heating and ventilation principles, and the sample's preparation (i.e., ground or not, ground particle size), size, and thickness. Third-party certified labs produce accurate readings, but those tests are conducted off-site in specialized vacuum ovens and typically take 6 hours or more. Hence, they're impractical for process optimization and control.

Figure 1. Ohaus

Most processors then compromise, using tabletop units without vacuum and reducing the measurement time to as short as 5 minutes or less, so their QC staff can take quick, frequent readings. Results can then be falsely low (because not all the moisture has time to evaporate), and error becomes more pronounced and variable with roasted nuts because their remaining low moisture gets harder and harder to remove. Of course, absolute accuracy becomes less important once norms are established. However, high variability still matters, and 5 minutes is a significant amount of time when processing thousands of pounds per hour.

Some processors use color analyzers (available from suppliers like Hunter-Labs), and some use manual color charts, slicing kernels open and scoring them against a color palette. The obvious negatives of the manual approach are that it is very subjective (and not repeatable) and time-consuming.

Measurement Methods Used

With various pros and cons, multiple technologies can be used to quickly (almost instantaneously) measure organic materials' moisture content, including sensors that use a microwave, capacitance, radiofrequency, and infrared principles.

We chose to use near-infrared radiation (NIR) moisture sensor (Figure 2) because its principle is best for low moisture contents, it's not affected by the sample's temperature or the ambient temperature, it doesn't require contact with the measured material, and only a very small sample (i.e., 5 grams) is needed.

Infrared radiation does not penetrate deeply, so the sample is ground up to measure its average moisture, but "weigh-dry-weigh" measurement also requires ground material. However, the NIR measurement is insensitive to the range of particle sizes produced by tabletop coffee grinders.

NIR moisture sensors illuminate the sample with near-infrared radiation that is absorbed in constituent-specific wavelengths. The moisture content thus affects the return signal's attenuation at water-specific absorption bands. Calibration is needed to convert the signal attenuation to a moisture measurement.

Figure 2. Near-Infrared Sensor

To do the calibration, we used a 3rd-party certified lab's vacuum-oven measurement combined with stringent sample creation and handling to compare the NIR sensor's readings to the lab's weigh-dry-weigh readings, producing a single linear calibration with a slope "m" and offset "B" (i.e., the % Moisture is equal to [m x NIR signal] + B) that we could use to measure the moisture content of any ground pistachio kernel (regardless of source, and whether in-shell or shelled, raw or roasted and salted or unsalted).

NIR Measurement of Roasted Pistachios

Then, to measure the color of the same ground pistachio samples, we used an optical color analyzer that we had previously developed to measure the color of seafood. This analyzer illuminates the sample with white light. It then analyzes the return light's color spectrum with a proprietary technique that scores the nearness of the ground sample's spectrum to that of two previously imaged standards, in this case, a very green pistachio standard and a very brown pistachio standard, with greener samples scoring closer to 0 and browner samples closer to 1.

Measurement Results

After deriving the moisture and color methods, we used them to measure a wide range of roasted pistachio market samples (from multiple brands and of various types, such as kernels and in-shell, and salted and unsalted) to gain insight into what values are typical and the degree of variability.

The results are summarized in Table 1. The average moisture ranged from about 2.3 to 3.5%, with an overall average of about 2.77%, and average color score from about 0.4 to 0.6, with an overall average of 0.52. The major brand samples generally measured moister and greener (averages of 2.95% moisture and 0.49 color score) than the "big box" private-labeled samples (averages of 2.47% moisture and 0.58 color score). Also, retesting of samples proved excellent repeatability, of typically 0.01% moisture and 0.02 color score.

To further explore the relationship of moisture and color, we plotted the market's scores on normal distribution curves created using the means and variabilities from Table 1, with one curve for moisture and one for color. Chart 1 shows the sample population's mean moisture was 2.77%, while its driest sample measured 1.76% and its moistest 4.00%.

One might expect the moister samples to be greener because they were presumably more lightly roasted, but that was not always the case.

Referring now to both Charts 1 and 2, sample 1 had a high 4.00% moisture and a low, green color score of 0.389, but although sample 2 was quite green (0.433), its moisture was surprisingly low (2.51%). Similarly, while sample 8 was both dry and brown (2.29% and 0.709 color score), sample 7, while very brown (0.701), was still quite moist (3.52%).

These results indicated that roasted pistachio moisture and color are not strongly correlated across the marketplace.

Roaster Test Matrix

To further investigate the relationship between roasted pistachio moisture and color, we conducted a matrix of tests with in-shell, unsalted nuts on our pilot roasting line (which integrates preheating, pasteurizing, roasting, and cooling into a single inline process).

At odds with the market testing results, where moisture and color did not correlate well, Chart 3 shows that on our roaster, the roasted moisture and color correlate strongly, with minor shifts due to the raw material.

Chart 4 shows that the raw materials' moisture and color were key determinants of the final roasted product. When their changes (roasted – raw) were plotted, the two datasets were nearly colinear.

And finally, Table 2 shows how on our roaster a given moisture and color combination could be achieved by different combinations of roasting times and temperatures, with a selection of an optimal recipe presumably then dependent on tie-breaking measures not considered here, such as product appearance, taste, texture, and shelf-life, etc.

Conclusions

Fast (seconds vs. minutes) and repeatable ways to measure pistachio moisture and color were developed and successfully used to gauge market norms and variability and conduct a test matrix on Laitram's pilot roasting line.
Results of the market assessment:
- Moistures ranged from 2.3% to 3.5%.
- Moisture and color did not correlate.
- Major brand samples were generally moister and greener than "big box" samples.
Results of the test matrix on our pilot roasting line:
- Moisture and color correlated strongly.
- Raw material moisture and color predictably impacted roasted moisture and color.
- Different combinations of roasting times and temperatures can achieve the same moisture and color outcomes.