Page 241 - Sensors and Control Systems in Manufacturing
P. 241

Networking of Sensors and Contr ol Systems in Manufacturing
                          for example, the hardware control volumes were denoted a, b, c, d,   201
                          and e, and software control volumes p1, p2, p3, and p4. The failures
                          within each of these control volumes were further categorized as:
                              •  Fault detected by an existing sensor
                              •  Fault that could have been detected if a sensor had been
                                 present
                              •  Operator learning phase problem
                              •  Failure due to manufacturer problem
                              •  Software logic problem
                              •  Repeat of a problem
                              •  System down for an extended period

                          4.7.8.1 Software Problems
                          It is often assumed that disturbances in the cell software control sys-
                          tem can be detected and evaluated relatively easily. Software diag-
                          nostics are common in most turnkey operations; however, it has been
                          shown that software diagnostics are far from perfect. Indeed, soft-
                          ware problems are of particular concern when a revised program is
                          introduced into a cell that has been running smoothly (existing bugs
                          having been ironed out). The availability of the cell plummets in these
                          situations, with considerable loss of production capabilities and a
                          concomitant higher cost. The frequency of software faults increases
                          dramatically when revised packages are introduced to upgrade the
                          system (Fig. 4.4). This is mainly due to human error on the part of
                          either the vendor or the operator. Two possible approaches to soft-
                          ware defect prevention and detection are:
                              •  Investigate software methodologies and procedures and rec-
                                 ommend alternative languages or more tools as defect pre-
                                 vention measurements. This is tedious and uses ambiguous
                                 results because such investigations are not based on data.
                              •  Analyze the problems that result from the current design and
                                 develop a solution for each class of problem. This produces less
                                 ambiguous solutions and is typically used to solve only immediate
                                 problems, thereby producing only short-term solutions.
                             To identify the types of faults that occur in programs, it is neces-
                          sary to know what caused the problem and what remedial actions
                          were taken. Program faults can be subdivided into the categories
                          shown next and restated in Fig. 4.5.
                          Matching faults:

                              •  Wrong names of global variables or constants
                              •  Wrong type of structure or module arguments
   236   237   238   239   240   241   242   243   244   245   246