Performance and reliability prediction for evolving service-oriented software systems

During software system evolution, software architects intuitively trade off the different architecture alternatives for their extra-functional properties, such as performance, maintainability, reliability, security, and usability. Researchers have proposed numerous model-driven prediction methods based on queuing networks or Petri nets, which claim to be more cost-effective and less error-prone than current practice. Practitioners are reluctant to apply these methods because of the unknown prediction accuracy and work effort. We have applied a novel model-driven prediction method called Q-ImPrESS on a large-scale process control system from ABB consisting of several million lines of code. This paper reports on the achieved performance prediction accuracy and reliability prediction sensitivity analyses as well as the effort in person hours for achieving these results.     

Development of a questionnaire to assay recalled liking for salt,sweet and fat

Liking for a sweet, salty or fatty diet may induce overconsumption of simple carbohydrates, sodium or lipids. Measuring overall liking of the corresponding sensory sensations contributes to understanding the determinants of dietary behaviours. However, no standardized validated questionnaire assaying these sensations is currently available. In the present study, we developed a web-based questionnaire, “PrefQuest”, which measures recalled liking for the following four sensations: salt, fat and salt, sweet, fat and sweet. PrefQuest included four types of items: (1) liking for sweet, fatty-sweet and fatty-salty foods, (2) preferences in the level of seasoning by adding salt, sweeteners, or fat, (3) preferences for types of dishes in a restaurant menu and (4) overall questions about sweet-, salt- and fat-related behaviours. A development study (n = 198) demonstrated that PrefQuest was feasible, well-perceived, only lasted about 20 min and that items were repeatable (overall mean intra-class correlation coefficient: 0.77, SD 0.08). Construct validity was then evaluated on a larger population sample (n = 47 803). The underlying structure within each of the four sensations was determined by exploratory factor analysis and then internally validated by confirmatory factor analysis. After a selection of the most relevant items, the salt, fat-and-salt, sweet, and fat-and-sweet scales exhibited a theoretically meaningful factor structure, unidimensional for the salt scale and with interrelated sub-dimensions for the sweet, fat-and-salt, and fat-and-sweet scales. Based on the fat-and-sweet and fat-and-salt scales, a fat model was also unveiled. For each factor, internal consistency as well as convergent and divergent validities were demonstrated. Overall, PrefQuest is an internally valid and original tool that can be used to assay recalled liking for sweet, salt, and fat considered altogether or separately as fat and salt or fat and sweet, and can be applied to large population surveys.     

Acceptability of partially dealcoholized wines – Measuring the impact of sensory and information cues on overall liking in real-life settings

In this study, the effect of partial alcohol reduction in wine on consumer expectation and overall liking was investigated in real-life settings (by a Home-Use-Test). By varying the wine in the bottle and by manipulating the information on the label, it was possible to assess the relative impact of sensory properties and information cues on overall liking.Standard Chardonnay and Syrah wines were partially dealcoholized from about 13.5% to about 9.5%. After bottling, these dealcoholized wines were presented once with information about real alcohol content (9.5%) and once with false information about alcohol content (13.5%). A third bottle with standard wine inside was also presented with real information about alcohol content (13.5%). A group of 194 French wine consumers rated the three wines at home for liking before and after tasting on continuous hedonic scales. Another group of 90 French wine consumers evaluated the wines at the laboratory.The same order of preference was obtained in both laboratory and home conditions but wines generally attained significantly higher hedonic scores in Home-Use-Tests. A strong segmentation of consumers highlighted specific consumer behaviors regarding expectation and overall liking of wines. Results emphasized the importance of both sensory and information cues in the development of overall liking by consumers.     

Poled Composites with Nd-Mn-Doped PZT Fibers Under Bipolar Electrical Load

Using the sol-gel process, Nd-Mn-doped PZT fibers were produced. The PZT was doped with 2 mol% neodymium and 1.1 mol% manganese. For characterization, the fibers were embedded in a polymer. The resulting 1-3 composites were poled with constant electric field. Strain and polarization were measured by applying a bipolar sinusoidal voltage of high amplitude. Instead of the expected shifted butterfly-shaped strain hysteresis, an asymmetric strain-field relation was observed. It is characterized by a rather linear region in direction of the poling field and an inflated region without strain switching for reversed polarity. Within the temperature range from room temperature to 80degC, the strain switching seems to be suppressed. Measurements of the piezoelectric coefficient at superimposed electric field prove the blocking of strain switching. Cyclation experiments with sesquipolar load show a pronounced linearity of the strain loops that declines after more than 2 times 10⁴ cycles.     

Uses of change-over designs and repeated measurements in sensory and consumer studies

The paper illustrates two statistical methods, the design and analysis of sensory experiments taking into account the effects of serving order and previously assessed treatment and the analysis of experiments with time repeated measurements.

Change-over design experiments balance both presentation order and carry-over effects. The proper analysis of variance allows the testing of these effects and the estimation of product means adjusted for carry-over effect.

Repeated measurements occur when groups are being compared over time. Either a corrected split-plot or a multivariate analysis of variance (MANOVA) with measurements at different times forming the variable should be adopted to allow proper tests for the time effect and for its interaction with the group effect.

The usefulness of change-over designs with an appropriate analysis and repeated measurement analysis will be demonstrated using sensory experiments involving trained assessors and consumers.     

Model checking C source code for embedded systems

In this paper, the applicability of model checking to C code for embedded systems is studied. The paper is divided into two parts. In the first part, 13 existing model checkers for C code are detailed and evaluated for their applicability in the verification of C code for embedded systems. A case study is presented that applied CBMC as one representative C code model checker to an exemplary microcontroller program. As a consequence of this case study, we decided to develop a new model checker for source code for microcontrollers, called [mc]square. It is described in the second part of this paper. We present the architecture and the peculiarities of [mc]square, and we successfully applied [mc]square to the same microcontroller program used in the case study.     

Impact of partial alcohol reduction in Syrah wine on perceived complexity and temporality of sensations and link with preference

Complexity is often stated as a mark of wine quality but there are no quantitative data supporting this assumption. The effect of partial alcohol reduction on the perceived complexity, temporality of sensations and liking of a Syrah wine was investigated.An Australian Syrah wine was partially dealcoholized using the reverse osmosis technique to obtain five wines with different alcohol content (13.5–8%). 71 French wine consumers evaluated their liking for these wines as well as the perceived intensity of overall complexity and eight items reported to be linked to perceived complexity in wine (familiarity, persistence, etc.) using continuous scales, anchored with pictures (Medel, Viala, Meillon, Urbano, & Schlich, 2009). In addition, eight trained panelists described temporality of sensations in the same wines with the Temporal Dominance of Sensations (TDS) method. The overall complexity scale was well understood by consumers with a significant discrimination of wines according to their alcohol content. Wines with higher alcohol content were perceived as more complex, persistent, strong and with many aromas. They were associated to more complex temporal sensory profiles, with many sensations that blend in-mouth along time.     

Bitterness and astringency of flavan‐3‐ol monomers,dimers and trimers

Intensity of astringency and bitterness of seven flavonoid compounds was evaluated by a time‐intensity (TI) procedure. Eighteen trained judges rated intensity continuously from ingestion, through expectoration at 10 s until extinction of the sensation. The seven stimuli included two flavan‐3‐ol monomers, (+)‐catechin and (−)‐epicatechin, three dimers and two trimers synthesised from catechin or epicatechin by condensation with (+)‐dihydroquercitin. As the degree of polymerisation increased, maximum bitterness intensity (I_max) and total duration (T_tot) decreased whereas astringency I_max increased. The monomers were significantly higher in bitterness at I_max than the dimers, which were significantly higher than the trimers. Astringency I_max of the monomers was lower than the dimers or trimers, although no significant difference was found in T_tot among the polymer classes. The bond linking the monomeric units had an influence on both sensory properties. The catechin‐catechin dimer linked by a 4→6 bond was more bitter than both catechin‐(4→8)‐catechin and catechin‐(4→8)‐epicatechin. Astringency was affected by both the specific linkage and the identity of the monomeric units with the dimer, catechin‐(4→8)‐catechin, being lower in astringency than either catechin‐(4→6)‐catechin or catechin‐(4→8)‐epicatechin. © 1999 Society of Chemical Industry