ML19209C854
| ML19209C854 | |
| Person / Time | |
|---|---|
| Site: | Crane  |
| Issue date: | 04/30/1965 |
| From: | Mertney R PHILLIPS PETROLEUM CO. |
| To: | |
| Shared Package | |
| ML19209C834 | List: |
| References | |
| TASK-TF, TASK-TMR IDO-17074, NUDOCS 7910180377 | |
| Download: ML19209C854 (32) | |
Text
_
_.1._.-..-
1 I
IDO-l'i 074 AEC Research & Develeprent Report General, Miscellaneous, and Progress Repcrts TID-k500 39th Edition Issued: April, 1965 FTETD EVALUATICII N:D CCITIROL OF E0I:I;EL EEHAVIOR By R. J. Nertney i
l i
PHILLIPS PETROLEUM COMPANY ThJ aab v
Atomic Energy Division 7D3M3037 Caerect Af ( 10 11 - 205 Idaho Operations Office U. S.
ATOMIC ENERGY COMMISSION 864 2
i i
AESTRACT This paper presents considerations in perfor=ing hucan factor l'
studies in a relatively unperturbed working environment and outlines certain specific programs with which we have had experience.
In establishing our system, we have chosen not to use individual mployee " worth
- criteria but rather to establish working roup in-c: as based on specific organizational objectives as stated by manage-ment. The ability of the organization to fulfill these objectives is sanpled by staff evaluators on a "did do" or "did not do" basis.
Reduction to simple dichotomy takes use of quality control tech-niques possible and provides useful numerical indices for purposes of evaluation and control of personnel behavior under field working conditions.
I I
I 864 262 i
l 3
TABLE OF CO:TfE2S Pace 3
l Abstract 7
Introduction Classical Methois 9
Objectives in Establishing a Control-Surveillance 11 System The Controllt d Plant 19 I.
Industrial Accident Control 19 II.
Operations Perse:~.ne ) Centrol 20 22 III.
Job Estinating Eel sbility.
IV.
Instrument Reliability.
25 V.
Operations Process Control..
26 Mathenatical considerations 27 29 Correlative Studies.
Extrapolation of Results 29 30 Conclusion 31 Acknowledgements References 32 FIGURES 16 Block Diagram - Industrial Safety Program Block Diagram - Operations Personnel Control 18 21 Occupational Injury Centrol Chart Operator Error Control Chart 23 Job Estimator Control Chart.
24 8b4 2b3
_5_
Introduction A constant problem in the operation of any activity engaged in by individuals or groups of individuals is insuring their proper inter-action with other individuals and groups. In the industrial environ-ment this takec the form of communication and control involving the various line organizational levels as well as " surveillance" groups within and without the organization. Historically, the " surveillance" functions have included diverse activities ranging from certified and internal audit functions to govern = ental surveillance of quality con-trol on food and drug ite=s.
In a typical industrial organization, one finds a great number of arcac which req tire surveillance and centrol. These range from situa-tions involving misuse of postage stamps and petty cash to surveillance and control prcgrams designed to prevent great catastrophes with attend-ant loss of human life and/or large scale destruction of property. In addition to int 7rnal control problems, there are more and more areas in which the potential interaction of an industrial plant with " civilian" population has resulted in the necessity for external surveillance i
grcups at one or several governmental levels ranging from municipal to federal agencies. Perhaps the outstanding examp' of the latter is the surveillance and control exercised in the hanuling of radio-active and fissionable raterials.
In our own activity, the operation of nuclear test reactors under contract to the Atonic Energy Connission, an exceptional num-ber of areas requiring both internal and external surveillance and control exist. Objectives which have been particularly important to us include, in addition to the usual ficcal and raterial control cain-tained in any well nanaged organization, certain elements specific to the nuclear plants which tust be maintained under extremely rigid con-trol. This is nececsary in order to assure ' hat an acceptable risk level with regnrd to nuclear ac vell ss the usual types of industrial accidents is being maintained.
The components of acceptable risk, or for that tatter, any form of " acceptable behavior", may be separated into three "P's"--plant, personnel, and procedures. These are not independent parameters and interaction of the three must always be considered. Excellent pro-cedures and rules are of no value if personnel do not know and obey them. Similarly, operation of a plant, which can be easily and safely operated by grr.duate engineers, may constitute an unacceptable risk when entrusted to the hands of high cchool graduate technicians.
Of importance in managing situations of this nature ic the crea-tion of a syste atic, objective framework for surveillance and control.
Too often surveillance and control sycterc are neither systenatic nor objective. To quote Dr. Sidney Werner, "The world today, q,s in the past, is threatened by prejudices of which racial, cocial and econc=ic prejudices are but a few. Equally influential, but less well recog-nized, is the prejudice of ' experience' derived from uncritical or 7_
bb4 264
uncontrolled observation, from the word of an ' authority', or from emotional bias. Fortunately, there are those who are villing to give ti=e and effort to seek out and correct such distortions of the truth".
(1)
The answer to the questien, 'How often do pecple malfunction to the extent of causing injuries which require medical attentien?", is, for example, often tabulated in pericdic status reports to management.
The seccndary questions involving variaticn frcn repcrt period to re-port pericd are seldcr answered. These include such questions as:
"Is an increase or decrease in frequency 'signiime.nu o/ u&es it merely represent an expected variation around ecce long term average value? If the variation is significant, what is the ' assignable cause' for the variation? What factors are important in peoples '
tendency to malfunction?" "If procedures designed to reduce accident frequencies have been affected, has the decided response been obtained from the organization?"
Another set of questions exists in connection with persennel per-formance in carrying out assigned jobs. Too often this sort of per-formance is treated with a wave of the hand generality or periodic completion of individual check sheet forts describing each e=ployees general characteristics. Such questions as, '{cv many and what sorts of errors does cparating crew nunber three take on evening shift?
Is there a significant performance diff rence frcn shift crew to shift crew? or Ecv vell do operating personnel conform to plant pro-cedures?", are too seldom evaluated and analyzed.
It is in an effort to answer questions of this sort that we have developed and are applying within our organisation certain " internal audit" or " quality centrol" progrc=s. These programs are designed to form a eccpatible cctplex relating the plant, the people and the pro-cedures under actual field perfor 1nce conditions. For example, a deviaticn in a tenitored process variable is related to what the people did do or should have done. Evaluation of the latter is in turn based on tne plant rules and procedures.
An unbiased central correlative agency is also maintained through which behavior of one working grcup functioning within ene surveill-ance--centrol program may be compared with behavior in other grcups within the same or different programs. One purpose of this, of course, is to detect behavioral influences outside a group's inmediate work-ing environment. Another function includes the study of related per-fronance deterioration or improvement frcn grcup to group or frem one class of misbehavicr to another. An example of the latter is the relationship of operational errors involving mismanipulaticn of equip-ment to personnel injuries.
d( 64 f63 P
i l
Classical Methode There are a number of classical cethods used in handling the problem: of sampling and assimilating the infernation required to j
control an industrial group.
i I
A.
Conventional Quality Centrol and Internal Audit These systems, which might both be classified as " internal audit" (or " quality control"), in the general sense, are highly l
developed. systems designed to sample organizational performance objectively.
" quality centrol" systems, and to an increasing degree " internal audit", also seek to d efine "significant devia-tion" in the monitored parameters. These systerm are applied ordinarily to financial and =aterial audit and to production quality control. They have received a minimal application in the formal control of line operating verk groups outside of these areas.
l B.
Supervisory Surveillance Direct surveillance by a superviscr is, of course, the classical and = cst videly used means of evaluating perfonnance of a working grcup. There are a number of practical difficult-iec which becc=e obvious when " intuitive" supervisicn ic used, however. This sert of surveillance becomes difficult because of lack of direct contact between certain levels of cupervision and the worker (particularly in the case of shift verkers), loss of objectivity through such phenomena as " halo" effect, tendencies to bypeas line supervisory levels (which aggravate the "typassed" cuperviscrs ' lack cf personal contact), and increased cc plexity of operation. Probably the ecst sericus defect of all is the tendency to spend a disproportionate a. cunt of superviccry atten-tien in preventing recurrence of the cost recent evidences of misbehavior. The latter can result in unnoticed deterioration of crganizational behavior in a different area. The result of all this is that a supervicor can becere " spread too thin" in tenitor-ing the behavicr of perconnel in a modern complex industrial situation.
It should be stated at this point and it will te reiterated frca time to tine that the " audit" syc ;en is no =cre intended to replace direct supervisory surveillance than conventional audit is intended to replace normal accounting procedurea cr, in engin-eering language, than a ficv recording cystem is to replace con-ventional surveillance and inspection of a pump.
C.
Investigating Bodies The "invectication" using unbiased personnel is a scst ecm-mendable effort to introduce objectivity into the study of mis-behavior and to prevent recurrence. The difficulties in uce of this cystem are several. b b 4, i
The first lies in the fact that in order to obtain individ-uals to " investigate" ebjectively, cne must sacrifice knowledge.
That is to sa/, the most knowledgeable individt als regarding the events being stud.ed are the people =oct intimately involved.
These people too cften, are not in a position to view events objectively and nuct te replaced by less knowledguable individuals in forming the " investigating" grcup.
The second problem is a statistical cne. The spectacular incident is selden a typical one by definition, and the basic day-to-day organicaticnal defect or defects which permitted the event to happen =ay or cay not be obvious.
The conbination of these factors makea it extremely diffi-cult for an investigating group to arrive
..t the basic cause of
=isbehavior. The result, as in the case of supervisory surveill-ance, is often the expenditure of a great deal of effort to pre-vent recurrence of a specific event of =isuehavior which is very unlikely tc recur. This lack of tendency to " repeat" may be be-cause of the low inherent probability of the event or because of the self-correcting tendencies of the organization itself. The investigating group can therefore go astraf in framing recc=:enda-tions which..ead to the correction of directly associated events and, more co=nonly, local environmental conditions associated with a specific accident but not the basic cause.
Another difficulty due to the statistical inadequacy of the incident investigation is the =isevaluation of individual menbers of the working grcup on a ccepletely inadequate basis. In addi-tion to the ;bvious short tern statistical sa:pling inadequacy, factor in judging per-
- 4-a there exists an important long
+a--
sonnel performance.
I. tany cacec, " accident proneness" appears to be time dependent and is a function of a time varying "off the job" enviren=ent as well as being age dependent. Both of these factors have been particularly well dencnctrated both in the case of autoncbile accidents and in at leact one Icng tern study of both occupational and nonaccupational injuries in an industrial con = unity. The diccucsion of " accident preneness" is beyond the f this paper, but such conceptc as " trigger episodes or s.~nts" the ability of individuals to learn to ecpe with external and psychological strescec without indulging in self-injury, and others offer logical explanations for the time variation in accident patterns.
The closely related probler of control likewise suffers cer-tain related inadequacico. Little need be said here except to state an analogue.
Frequently, personnel performance ic controlled using much the care philosophy as an individual who would attempt to drive hic autencbile by " feel" or "intuiticn" withcut 5 function-ing speedc=eter. Too often, 80 miles per hour " feels" like 60 be-cause of road condition, state of mind, etc.
Similarly, factorc involved in performance evaluation by " feel" ccre into play to render control ineffective when based on purely cubjective reactions dr 6 9 4
d6-
4 At thic point, to a gross impression of personnel perfemance.
tf we might do well to again refer to Dr. Werner's statement quo e in the introduction of this paper.
Objectives in Establichine a Centrol-Surveillance System The criteria for an effective infomation and control system in-volving both per:ennel and equipment are several.
The sycten chould supply numerical infomation reg'arding perfor-This infomation should be of a quality and quantity whichThe information sh permits use of appropriate nathematical codels.
nance.
d te in such a fem that it may be used by supervision to guide cembers of the population being con-perfomance.
of suitable information to thetrolled in order that they night i= prove their pe l
protect themselvec from faulty evaluation.
cation should te available in appropriate form, including proper
- nel, l
guides to interpretation, for the use of staff and service person top level management and external agencies, such as governmental sur-In the case of tany, if not most of the surveill-h ogran ance groups, krowledge of the general criteria governing t e pr veillance groups.
can constitute " sufficient infomation" for their use.
In connection with the control function, the progren should:
Provide quantitative or at least semi-quantitative evaluation of perfor: ance to pemit evaluation of 1.
response to control.
t Focts supervisory attention on problen areas.
2.
Maintain a naximum autonomy of working groups and line 3
supervicors.
Pemit statement of organizational objectives in terne 4.
of quantitative indices.
strengthen the line organization concept by directing the attention of cupervicion and management to the specific 5
degree of detailed perfomance analycic and perso Aid in control of cpec' acular misbehavior by providing infomation regardin; cisbehavior patterns and through 6.
control of " routine" cisbehavior.
The proposed method of attsch is through a femalized uce of information and control " loops" which already exist in tcat organica-The basic loop componente in a " personnel" loop ith concist of, (1) the, people or groups of people who are charged w tional structurec.
l performing a job and (E) the st.pervicor and staff vno are recponcib e 864 26
for supplying instructions and seeing that satisfactory work is being done (co==unicati in and control). In the =ost ele =entary loop, the
" loop" itself is for:ed frc= a working " block", a supervisory " block",
an arrow indicating control and an arrow indicating infor ation or work sa=pling. Similar loops applicable to =echanical syste=s =ay be constructed by replacing the bicch representing the working perconnel by a block representing verking equip =ent.
The loops are coordinated through line and staff organization to insure ec=patibility of person-nel, equip =ent and objectives.
Before examining specific 1 cops, let us outline a few of the considerations involved in setting up a ec=prehensive control progrc=
providing for its own " internal auditing". In the first place, we
=ust decide wheth?r a perfor:ance factor is best studied on a group or individual basis. This depends on a ncsber of things. These include:
1.
The steps en the organizational chart between which the prograr is to fanction. A plant =anager might be inter-ested in the gross length of time required for his organ-1:ation (group) to cc=plete certain jobs which involve all crafts. This interest could propagate itself down the organization and result in a progra= set up betvcen veldint forenen and the individual craft velders to evaluate and control ti=e spent by each of the velders in perfor=ing his job in the field as well as programs of vider scope designed to deter =ine how various shops and groups perfor= as units.
2.
The scope of the progra=.
If, for exa=ple, one wished to evaluate the perforrance of individual design engineers to developing workable designs, a co=parative progran might be set up to study the ability of the fabrication shop as a unit to satisfactorily construct work done by each individual =enber of the group of designers.
If, on the other hand, one wished also to evaluate perforn-ance within the fabrication shop, cc=parative perfor:-
ance of velders, fitters, forc=en, etc., =ight also be studied as individual units.
3 The state of the progra=. It is often desirable to screen the factors involved in attaining an organica-tional objective en a coarse scale. After the effects of influential cross factors are determined, cne can proceed to a finer structure involving s= aller work groups and = ore subtle factors in order to apply more specific corrective action and control.
The three questions involved in perfor ance sempliag are:
1.
What to sa=ple?
2.
Who to sa=ple?
3 Who should do the sampling?
h6/j
}fg -
i There is no "ccckbook" answer as to what to semple. The best answer to thic question is a counter question.
"'Inat are ycur organi-zat'.cnal objectives?" As ctated previoucly, the infomation collected must be equivalent to or directly related to the organizational object-ive being controlled. This is the first and forc=oct criterien. Sample size chculd be great enough to be statistically significant and repre-centative.
" Depth of cut", ic i=portant if the parameter campled is one that varies in importance. This refera to acceptance of data ite=s which are related to, but are of lecc importance than a sm'l er subgroup of events one has a strong desire to control. Obviously, increasing the " depth of cut" dilutes infomation regarding the items of greater impo--tance. If one atte= pts to collect information on
" Crime" in the United States by performing its: counts on all offenses ranging frc= =urder to violations of lawn watering ordinances, he vill not learn =uch about serious cri=e.
On the cther hand, he might learn a great deal about future sericus crimes by studying cc= temporary nar-cotics addiction cr juvenile petty crimes which may be logical precurs-oro of serious crime. Likewise, studies of 'less serious " offenses give general information regarding " national = orality and attitudes "
which =ight influence future serious crime rates in a less direct The diluting influence of a large " depth of cut" may be
=anner.
=inimized by use of weight factors based en Scportance.
In the example above, ten violations of lawn watering crdinances might be considered to be " wor %" one event of disorderly conduct. That is to say, the law is broken.n both cases but the event of disorderly conduct is considered by the evaluator to be, sociologically speaking, " ten times as bcd" as the violation of the vatering crdinances in setting up his sociometric scales.
Si=iltrly, the " depth of cut" into cperational behavior deter-
=ines the specificity of tLe infomation being obtained. Again, ther; is no standard rule; and eaah situation must be handled logically according to the objective desired.
The questions of objectivity, reliability, and consistency arise in connection with this type of program. Objectivity should be maxi-
=1 zed in two ways. The indicea should be dete mined in such a manner as to =aximize objectivity, and the individual (s ) who does the sampling should be selected and trained to incure the highest possible degree of objectivity. Often the index selected must, by its nature, te highly subjective. In a situation of this kind, one should remember that the program has not added a new problem. If the subjective inder is !=portant, it vac being evaluated intuitively pricr to existence of the campling program. In thic connection, it is important to real-ice that consistency in sc=pling vill often yield useful info =ation even if certain types of unreliability exist. It was demonstrated locally in one experiment that although two evaluatorc differed signif-icantly in abcolute numerical evaluation of a highly subjective factor, they were very consictent in evaluating changes in the factoI' from
=cnth to month and frc= cne numerical level cf =icbehavior to another.
I' Ulis is similar, in the infomation sense, to having one evaluator measuring te=perature en a Fahrenheit scale and the other on a Centigrade.
! 804
,0 1
I In mking the decision as to whom to ss=ple, one has three choices:
A representative sample of the controlled group.
1.
All of the controlled group.
2.
A larger population including individuals outside the controlled group (extended population).
3 a
The first is often done in pilot studies or in a functioningThe extended p program for econc=ic reasons.in detecting causative effects outside the working en group.
One may also stady sub-grcups of the controlled population.
Generally speaking, once a progras is operative, it is visest to y
establish control on the entire population together with as man The functional sub-groups as the data vill statistically per=it. rea j
the groups, it requires very little additonal effort to process be obtained.
sub-groups; and a great deal of infor:ation ray often One has only once to reprocess a great deal of accu =ulated data to f pro-add, say, a vorking shift breakdown to realize the wisdem o iginal gram::ing "tco much" infor nation on fine structure into the orAn eve m
extraction due to oversights in setting up a sampling progra.
data readout.
Another decision which cust be =ade, is whether the sa pling function should reside within or without the industrial groupA similar sit being controlled.Here, one may learn about an individual by asking him h ir questions about himself or by studying the indivi5aal's be av o psychology.
ible,
as seen ar.d interpreted by friends, neighbors, or, if possThe consideratio One must determine how the individual's ( or controlled trained observers.
(or itself) inter-work group'c) core intimate knowledge o' Welf the care.
in acts with his (or its) lack of objectivity and desire to appearHere, th a good light".
lities and objectivity built into the camplin6 procedure, the personatech the se=pling, and the particular organizational situatien.
The final dccision, which cust be =ade (and continue to be made f
nce. One as the program centinues), involves the use of past per onna in past must decida whether average past perfor:ance and variation ce.
He performance are suitable criteria for " satisfactory" perfcr:
fro:
past performance.
Two statements of variation which usually develop as those of "signi-prime importance once the organization is stabilized are i tion". 864 27
The former generally requirec investigation,.rhile the latter usm717 requires prompt corrective action.
It is worth repeating at this point that there are only two criteria in index selection. Firct, the index chould reprecent an organizaticnal objective and recond, it should te expressible in a quantitative tanner.
The second requirement will often diccourage the potential progran user before he starts, but it has been our enperience that if one exercises sufficient ingenuity alnoct any " objective" can be expresced quantita-tively and can be associated with appropriate cathematical =cdels.
The index selected tells us "what" is happening.
It is often desirable to break the "what" down into sub-c.ategories or, more often, to find out "why".
This leads to " factor anaJ ysis ". The purpose here is to provide a finer structure of infomation to facilitate understand-ing and to aid in selection of appropriate therapeutic control action.
The " factors" may take two forms. They may Le tangible factors lead-l ing to specific countermeasures, or they cay be intangible factors l
leading to a supervisor's general knowledge of his people.
Exampler of the former might be sub-tabulations of misbehavior due to " lack of training" or " lack of proper equipment".
In these cases, therapy is very obvicus. The latter is exemplified by cub-tabulations involving misbehavior due to such cauces as " absent mind-ednes s " or "has te ". Here, specific corrective action is much less obvious.
The factors may be exhibited in the fo= of average distributions, or changes in events ascociated with the various factors may be studied as a function of total number of events of nisbehavior. The latter forn of analysis permits one to determine whether variation in a single factor predominates au general perfo m ance deteriore.tes or inproves.
This whole procecs ic not as complicated ao it night sound and is 4
best illustrated by example.
I i
Case No.1, Industrial Accidents Lec us consider the typical plant safety structure illuctrated in Figure 1.
This illuctratec a line organization one of whose objectives is to prevent industrial accidents. Frhre recponsibility rests in line organization. A ctaff Croup (or individual), reporting to management, functions in an advicory capacity to ascist line organization in fal-filling the objective.
The dotted lines bdicate info mation trancnittal, and the solid lines represent control. In thic case, the staff cafety engineer in cooperation with the staff cpecialict control group celecta an appro-priate control index, such as firct aid caces, minor injurice or plant condition sm: pled by incpection. This index is campled from the field.
'5-864 27
~
i l
Top Management
\\
\\
\\\\
\\
N u
,r l
ir I
\\
Intermediate Safety l,+_ _ __, S to f f Cont rol Supervision
/ Staff Group Specialists
,/
n
/
I
/
l 1
/
<r l
Field p
l h
Supervision I
I I
I Field
,_______a Worke rs Sampling Information
In f orma t ion o c' PPCo - A - 4 868 g
'g Fig. 1.
Block Diagram - Industrial Safety Program u
t h
w 8O
After process".g the data the safety engine ~er uses the infomation gathered to guide him in his educaticnal programs and in evaluatien of plant systems. The centrol data are also fed back into line organ-1:ation in the fo= of control charts to guide supervisors in super-visory control. Examination of this case reveals at once that in a working system, not one, but tany feedback 1 cops usually exist.
It shculd be noted that the ctaff systen control specialists are not involved in sanpling, infc =ation feedback, er administrative control.
Their sole function is to assist in fc=ulating the program. Details of index selection, campling and control of a particular safety pro-gram vill be discussed later.
Case No. 2, Percennel Control This is a more sensitive type program in that a = ore serious organi-zational bypass threat exists. Ccnsider again a typical organization as shown in Figure 2.
In this case, the objective is to mininite operational errors in the " area".
The centrol systen functions as follows:
--5.The staff assistant samples operational errors cc=mitted within the working groups. Since thic is a g: oup control progran at area level, sanpiing is on an ancnymous basis with regard to the specific individual ce==itting the error. The detailed description of the error is stated, however, Infomation is fed back to the area manager who exercises control through the group supervisor. Inforna-
=ation is transmitted to superior manacement only at the area nanager's discretion. In this program, a very strong primary loop exists, but first line supervision has been included in the controlled working grcup. Progran :entrol is exercised from above only through the area
=anager via the first line supervisica to the working group. Again, there is no sampling, transmittal of infomation or control function associated with the staff centrol specialists. Adninistration of the progran is ccepletely in the hands of line superviscra and staff per-sonnel who were already respcnsible for achievement of the objectives covered by the program.
The control system for instrumentally monitored mechanical system parancters functions exactly the same way. The instrument data are senpled by the area staff assistant. Data are reduced with the assist-ance of staff centrol and are fed back to the area manager. Regulations
- f based en "significant deviation" are developed by the area manager, and control is established through these regulatives. Adequacy of the regu-latives in influencing persennel behavior is dete m ined by the personnel control program. This closes the loop and insures ec=patibility of l
objective, equipment and personnel behavicr within the scope of those activities included in the program.
l 864 U4 I
! l
Top Manage men t i
i l
I u
l Intermediate Management t
4 l
l v
l I
- -
- Staf f Control Operations Area Manager" S taf f Ass,t ]
Specialists t
I j
l i
T ~ ~ ~~ ~' ~ }'
\\
I I
l l
l I
e
+
v 4
<r
+
I Group Group Group l
Supervision Supervision Supervision l
t
+
t i
I I
I I
l I
l l
I u
y
+
u e
l s
1 Working Working Working I
I Group I
Group Group I
I L__________________J
~~~~~~
PPCo. - A - 4 8 67 Con t rol Fig. 2.
Block Diagren operations Personnel Centrol 864 275
1 The Controlled Plant Cutlined in this section are several programs with which we have had experience:
I.
Industrial Accident Control 1.
Objective: To reduce the nunter of industrial accidents in our plant and obtain general information regarding our personnel.
~ dex: First aid treatment for occupational injury.
2.
_n 3
Factors Analyzed: Welve accident " types ", seventeen
" agencies ", eight " unsafe conditions ", twelve " unsafe acts ", seven " personal" factors.
4.
Population Groups: Entire plant, five sub-groups (two operating groups, two maintenance groups, "othe rs ").
5 Data Presentaticn: Control chart with upper control limit based on two-year previcus performance and up-dated once per year.
6.
Sampling: Plant nurses.
T.
Factor Analysis : Independent analysis for each case by two safety engineers.
8.
Data Distribution: To all line superviscrs.
9 control: By line supervisors.
10.
Corrents: Care must be exercised to insure that the progran objective, as understood by the controlled population, is injury prevention not reduction in reported injuries.
j Correlative studies performed locally show that our first aid cases correlate significantly with our more serious injuries.
I 11.
Figure 3 demonstrates the chart for used for industrial accidents.
Data are normalised on total manhours worked and control limits are cet at 95% confidence. The statenent 955 confidence means that, based on previous experience (in this case, two years), the gambler's odds are 20:1 that an assignable cause may be found for a point lying above the upper control limit.
It should be noted that the control limit is probably not strictly applicable 8L4f4
~, n
.'. /
_lo.
I because of the downward trend starting in January, 1961, when the plant vent "on control". Even though the utility of the upper control limit var comprcnised by rapid performance hrprovsncnt, the rule for setting and updating control linits stated in (5) ateve vas maintained. This was dcne because nor- seasonal variatiens n17 be experienced which exhibit changes of the order of cur " improve =cnt" which is just ner becoming significant in 'he statistical sense. This chart for the entire plant is typical of the charts used for the personnel sub-groups.
l l
t II.
Operatier.s Personnel Control i
1.
Objective: To control errors of omission and con =ission by nuclear reactor operating personnel.
l 2.
Index: Errors committed by operating personnel.
3 Factors Analyzed: Infor.al separation into " omissive" and "conmissive" errors has been made. A four factor ana'yais based on "cmissive", "connis s ive ", " manipulative",
and " documentary" is being studied. The "documentry" i
error involves failure to log or record data in the
{
cmissive sense and improper logging or recording ec -
missively. Separation on " running" and "shutdo. n" reactor aspect is being piloted for one reactor.
Chronological shift (day, evening, night) is also being I
piloted.
i f
I h.
Population Groups: Entire operat.4ng branch with shift crew sub-groups.
}
l 5
Data Presentaticn: Control chart with upper and lever l
control Itaits for each shift and for entire branch.
Tabulated errors.
l 6.
Sampling: Operation: branch staff engineer (Operations Coordinator's effice).
I i
a T.
Factor Analysis : Operations br1nch staff engineer (Cy:raticns Coordinator's office).
i 8.
Data Distribution: Control charts and tabulated errors to branch operations superintendent--elsewhere at his discretion.
9 Centrol: On shift basic by operatiens branch superin-tendent through the shift supervisors.
10.
Comments : Procram very effective.
d,.6 4 s?/.
OCCUPATIONAL INJURY CONTROL CHART TOTAL AREA MTR -ETR ARE A MONTHS OUT OF CONTROL 20 E
CO
- 10
-20
-30 l
JW M8.06
-50 86
- 66.74
-5510
- 54.82 JAN FED.
- APR, MAY JUNE JULY AUG SEPT OCT NOV.
DEC.
~
/
MONTHS OUT OF CONTROL g
20 f
\\
JANUARY g
10 JUNE
\\
/\\
9 N
/ \\
^
l f) ~l0
\\ /
\\
/ \\
/
]8 N/
1
-40 a
i d \\.)
N NOTE: POINTS ABOVE ZERO IN DIC AT E E XCE SSIVE NUMBER OF INJURIES Co Fig. 3 Occupational Injury control chart
I 11.
Fi,;ure 4 indicates a typical centrol chart used in evaluating performance of operating personnel.
Ihe index here is one based on a set of " operating errorc" developed by the operatienc superintendent and the staff evaluator.
'"hece " errors " are selected in cuch a manner as to constitute a " representative sample" of operating proficiency. Similar charts are maintained for each operating crew and for the entire plant.
III. Job Est'cating Reliability I
1.
Objective: To evaluate and control reliability of ti=e ectirates in connection with field maintenance verk.
2.
Index: Ratio of tire estimated to perform a job to the time actually required.
3 Factors Analyzed: Estimator, originator, craft shop.
4.
?cpulation Grcups: Estimator, criginator, craft shop.
5 Data Presentation: Tabular and graphical.
6.
Sampling: Maintenance tranch staff personnel.
7 Factor Analysis : Factors are unambiguous and require only tabulation.
8.
Dsta Distribution: Relative performance of ecti=ators and craft shopc to Maintenance branch manager; relative performance of originators to criginator branch canager.
9 Control: Ey appropriate branch nanager.
10.
Co=ents : Program in proccss of development.
11.
Figure 5 demonstrates the chart form used in analysis I
of deviation of time required to cceplete a craft job j
from the job ecti= ate.
j A chart of this type is prepared for each job estimator from which two piece: of information are extracted.
Firct, hic overall tendency toward 1cv or high eaticatec may be determined in terns of the clope and intercept of the linear appropriation to hic performance (the center line). This line for an eatinatgr "vithout bias" would pacc through 0,0; 1,1; etc. at 45 The cecond piece of infonnation is obtained from the variance 864 279.
0 7
)
4 1
R 6
P 0
N O
4 1
l
(
A.
i C
o 9
U
{i C'
6 F
O 1
P F
R R
)
~
E D
8 1
W SS 6
O 1
(
P 1 1 N
( (
a P W R U O O
~
t
{+
6 T HE 1
/
SS R
=
t DR r
S S P a
6 hC 6 R 1
E l
B o
M r
t q
U n
N o
5 C
6 EL r
1 C
o Y
r C
r E
R r
)
i T
o t
4 t
M a
P 6
1 re 2
p O
(
)
)
1 3
1 3
3 4
7 6
f
(
1 g
(
i S
F
~~
S T
T I
I
)
D M
M
)
SS I
2 t
L I
L 6
1 1 1 L
L
(
(
O O
R R
e T
T N
N O
1 O
C J
5 C
4 R
R E
E W
P P
O L
U 0
6 1
0 8
6 4
2 0
1 U' > 0 x mJt 4LL 2
mh vOD CC gmW8
100 i;
90
- O.120 + 1.056 Est Act
=
S = 0.9 563 i
80 70 l
l 60
/
_J<
D w 59
/
/
0
/
1
/
/
/
/
40
/
30
/
20
/
./
//-
.K7
/*
/
Ppco a-ases O
10 20 30 40 50 60 70 TIME X = ESTIM ATED Fig. 5 Job Estimator Control Chart 864
?81
.s.
i experienced in comparing actual time consumed with time esticated. The two outer lines are adjusted in such a canner that the odds are 19:1 that an estimated-consumed time pair will lie between the lines (95% confidence)
The " spread" of these lines provides a nessure of the estimator's consistency in providing reliable estL:ates. A vide envelope, even in the absence of bias, indicates poor performance.
In fact, this is the best indicator of estimator per-formance, since even those estir2 tors who are extremely erratic often show little bias '. '.. ri'ssge values (over many jobs) of estimated-consumed time are studied.
It should be noted that nothing is indicated here as to whether a job should take core or less time. This representation indicates only the ability of the esti=-
ator to anticipate the amount of ti=e required. The philosophy here is that the estimator cust esticate in terms of the field working groups available to hi= and must acec==odate to their performance. This is true whether the field personnel are more or less capable than " average cen".
This same representation is used to evaluate perform-ance of cther personnel associated with the jobs.
In-verting the statement regarding the 95% confidence levels, we may state that the odds in favor of our findings an assignable cause for a point outside the envelope are 19:1. We cay, then, analyze the perform-ance of, say, work crew forecen by associating points outside the various esticators envelopes with job fore-men performing the work.
Tue indication of poor perfornince on the part of the foreman consists of generation of a disproportionate number of outlying points. Similar analysis can isolate difficulties in craft shop, design shop, originators, expeditors, etc.
The ability to perfor:
this sort of analysis is, of course, dependent on the degree to which the different estimators, shops and
'other personnel are intermixed in their work assign-cents. That is to say, if one esti=ator always estinates for the sa=c foreman and the same working group, it is impossible to separate forecen, work crew and estimator responsibility for outlying points by this method.
s i
IV.'
Instrument Reliability 1.
Objective: To study and control reliabiltiy of pro-cess instruments.
2.
Index:
Instru=ent performance under field and/or bench conditions.
9 864 g
9
3 Factors Analyzed: Instrument manufacturers and models.
h.
Population Group: Sample instruments.
5 Dcta Presentation: Tabular I
6.
Sampling: Maintenance branch staff.
T.
Factor Analysis: Factors are unambiguous and require only tabulation.
8.
Data Distribution; Maintenance branch manager.
9 Centrol: Maintenance branch.
i 10.
Comments: This program is on a " call" basis.
Ins tru-g ments are evaulated as requested by the branch manager.
V.
Operations Process Control 1.
Objective: To define " sir.ifievit deviation" and "naximum tolerable deviat> n" 1.q instrumentally monitored process parameters to assist the Operations branch superintendent in establishing process control.
2.
Index: Monitored system parameters.
3 Factors Analyzed:
"Trta" process variation and instru-ment malfunction.
h.
Population Group: Plant instrument systems and the process system itself.
5 Data Presentation: Control limits cn system parameter value and on variation in parameter. The latter is exemplified by instrument or system " noise".
6.
sampling:
Instrw ent charts.
7 Factor Analysis: Operations branch and maintenance branch staff engineers.
8.
Data Distribution: Operations branch superintendent.
9 Control: Operations branch superintendent.
10.
Comments : Program is in formulative stage. Proposed control vill involve variatien la individual paraneters and simultaneous variation in several parameters.
, 864
?g3 i
Mathematical Considerations It is beyond the scope of this work to present a course in be-havioral statistics and such a presentation could, indeed, be very unwise, for in this area a little knowledge is often a very dangercus thing. A few concents should be made hcVever.
Probably the most important fact to note is that the field of human behavioral studies is no place for the amateur statistician.
In our own case, we are fortunate in being able to utilize the ser-vices of our Operations Analysis group which is. composed of profes-sional mathe aticians with speciality kncvledge in the field of statistics.
The problems which one encounters in dealing with the human factor in the field are many. There are of necessity many highly subjective factors which force us to specialized mathematical models when ve attempt to evaluate "tehavior".
Our sitatuien is further compromised in that extreme care must be exercised to present upset of the primary organizational objectives. The latter consideration results, at best, in a pocrly controlled " experiment" and at vorst in serious confounding and obscuring of parametric behaviour data extracted frc= the multivariable system.
The first numerical analysis which we may perfonn consists of cerely counting the people. The next step might involve assigning identification so that we =ay distinguish one individual or group of individuals free another. These are relatively simple cathematical tasks, although ::.any clerical-mechanical problems arise in a large complex organization.
The more difficult problem is the quantitative evaluation of the various aspects ef behavior. In this situation we are able to reduce information of varying degrees of content, viz. :
i i
I 1.
Yes or no information. A =an or a group of men do or do I
not possess a well defined property (here " property" is f
used in a very general sense which includes, for example, the propensity to co nit certain acts).
s i
2.
Simple rank information. A =an or a group of men possesses more or less of a property than other can or groups of men, l
but we have no quantitative measurement of the property.
An exa.nple of this sort of information is that obtained by arranging a group of people according to height.
3 Quantitative differential infonration. In this case, we i
know "hov nuch" nore or less of the property. This situa-tion vould exist in case (2) above if our =en were, standing on a level surface behind a vall, and we could =casure dif-ferences in height with a reasuring device but could not see their feet to determine their total height. 864 284
4.
Ratio information. In this case, ve have sufficient infor-
=ation to determine "hov =any times as much".
To return to our exc.=ple, we would have, in this case, the height of our people in feet. We can then state that the 6' tall individ-i ual is o/5 as tall as the 5' tall person.
Each of these classes must be processed according to the appro-priate mathematical methods.
In addition to infocation class, there is a further problem which is extremely serious if one attempts to " cook book" statistical procedures. This has to do with the statistical properties of the data.
In the field of descriptive statistics and in tce use of statistical inference, certain postulates exist in the cathematical codels regard-ing the distribution and quantity of data.
If these postulates are not true in the hu=an-mechanical system, the inferences drawn are likely to be false. In some cases, the relationships with the math-e=atical models are identical with those encountered in conventional g
quality control; in sc=e cases, they may be quite different. A very important difference, for example, is in the population homogeniety.
This may be exemplified for one type cf difference by analogy. If a component in a soda cracker packaging machine f ails in such a manner as to randomly break a considerable number of crackers out of the total throughput, one cracker entering the cachine is no more or less likely to be broken than any other. That is to say, no cracker is " accident prone" with regard to the defect in the machine. On the other hand, if one introduces a machine into the plant which tends to skin the operator's knuckles, he would probably find that each operator vould have his own probe.bility of in, jury. If this effect were strong, it would be impossible to use " cracker" statistics to study the human population.
In addition to the statistical problems, the loop feedMck char-acteristics must be considered. One of the evils of the " free-running" organizational system which one vill often encounter is a feedback characteristic which results in little or no response to minor items of =isbehavior and extremely violent response to " serious " ite=s.
This is analogous to an improperly designed steering system for a motor vehicle in which any response on the part of the driver vill drive the vehicle frc= curb to curb. In both cases the result is an erratic course. This organizational defect is especially i=portant early in progran formulaticn when " pilot" stati:; tics, i= properly used can result in a great deal of organizational turmoil.
There are many formidable problems associated with assuring that the eystem being analr :cd fits the mathematical model, that rigorous
~
methods are being utilized within the mathematical model and that results der. =d frc= the model are being correctly related back'to the system. This means that team methods are often most appropriate in initiating programs involving the surveillance and control. The services of the mathematician are, in our opinion, essential to the Ohd [8}
8 success of our programs. Our working tesm usually consists of an engineering staff assistant from the group within which the program originates, a mathematician, and a panagement technical staff assistant.
Correlative Studies Once a nunter of human and/or plant parameters are under obser-vation in one or more working groups, it is desirable to maintain a central data analysis point at which the indices maintained by various working groups may be compared and evaluated on an intra-or inter-6roup basis. This is not an external " control" function but serves to detect interactions and cct=cn factors to assist the line supervisors involved in the individual programs. At the present time, we maintain, on a common calen.iar time base, the Operations personnel performance control charts, the first aid control charts, the serious injury in-dices, and an index which indicates the number of personnel reporting to our dispensary during working hours for medical treatment because of flu, colds, and other non-occupational complaints. The purpose of the latter indices to determine whether increased incidence of " sick-ness" in the area impairs operating efficiency.
Extrarolation of Results In many cases, we are interested in controlling spectacular
=11 functions or events of misbehavior which are not of sufficient frequency to admit statistical control. A few cctments vculd be made regarding the projection of the type of control described here
' ~ ~ ~ - - ~ ~ ~ ~
to these situations. In the first place, many of the events controlled in our programs might be considered precursors to the more spectacular type incident either generally or specifically. The concept of the
" accident triangle" is videly accepted in safe'.y engineering. That is to say, frequency of exposure to accidents is closely related to fre-quency of minor accidents and in turn to frequency of serious accidents.
This means that or.e might consider a program based on controlling the r
statistically numerous events of relatively minor consequence as a barrier which lies between acceptable behavior and behavior patterns which lead to many (but not all) of the events of core serious conse-quence.
Even if one does not choose to accept this concept, there are certain questions which come up in determination of the probability of the spectacular mishap which may be better answered if one main-tains this type of program. These include, for exsc.ple:
1.
Ecv vell do the plant personnel conform to " advertised" rules?
2.
How often do people indulge in " foolish, unpredictable" behavior?
l 3
What types of misbehavior are observed?
- 864
?g6 j
4.
Why and when do people misbehave?
Conclusion In conclusion, we might summarine the benefits which we have observed and which we expect to achieve in operation of our control programs :
1.
Discussions and meetings in connection with selection of control indices result in a healthy re-evaluation and re-statement of organizational objectives.
2.
Objectivity is increased in performance analysis.
3 A " smoother" organizational control system is achieved through correction of events of misbehavior before they result in consequences of major importance.
Typically, two effects are noted as control is established:
a.
The variation in the index becomes less from time period to time period.
b.
Index performance is improved.
1 4.
Supervisor--employee communication is i= proved and put on a more objective basis.
5 An objective documentary background of perfornance is established to protect the organization from " incident hysteria" 1.e., over-reaction or improper reaction to individual spectacular events of misbehavior.
I 6.
Gradual deterioration of performance may be detected.
T.
Careful study of reasons for calperformance is encouraged on a continuous basis.
j i
8.
Supervisory attention is " focused" on problem areas.
l 9
Chronic difficulties which have beccme "part of the local scenery" are better detected.
I i
10.
Surveillance groups outside the operating organization may i
study a " system" rather than indulging in the usally futile i
attempt to evaluate individual performance from "outside".
11.
The programs assist in firmly establishing the usually pcorly defined terms "cignificant va"lation" and " maximum folerable variation" in connection with internal and external adminis-trative control.
l 864 287 -
Co==ents regarding detricental effects anticipated prior to initiation of the programs and not observed include:
1.
Working group morale is evidently not comprised by an active progran and can actually be improved because of a feeling of " fair evaluation" in the controlled group. Care cust be exercised in introducing the program, however, and orienta-tion and conditioning teetings with personnel are essential.
2.
There need be no loss of autonomy for the working grogps who do their jobs well. If anything, g _; perfor=ance may be reva-ded by increased freedom from expilert external control.
3 There is nc loss of autonomy for the supervisor as the pro-gra is outlined here. Control indices are selected, sampled and controlled by the supervisor hi=self subject to pre-existing organizational objectives. The system is not in-tended to be a " cook book" shortcut to effective supervision but requires even core supervisory acumen than a free-runnir4 systen.
I 4.
There is soue risk in imposing this sort of control over a vide range of activities (e.g. " operator proficiency") that only the index sample of perfomane : vill be improved. In order to mini =1:e the possibility of this occurring, one should:
a.
=ake the index sa=ple items as representative of the ettire range of activities as possible.
b.
use a number of very specific index items to set the
" tone" of the index and include general non-specific items to broaden the coverage, c.
Obtain samples of non-index performance periodically.
I 5
The financia?. cost of the progra is essentially nothing.
t A properly selected behavior index =ay usually be sampled by personnel who are already responsible for surveillance and/crcontrol. The progran cerely fe malizes sampling
-and control in connection with Laportant organizational objectives.
i Acknowledge =ents As the reader night judge frc= the diversity of our samples indicating areas in which we have worked, a great many people other than the author have participated in organizing and administering
[
~
our programs. This is more true than is apparent in reading =y account, and in fact, my only excuse for authcrship is the fact that j
I have had contact with all of the programs.
864
?88 1
3-i f
h6e N= w h g.
g
,, g i
me g
Mr. R. L. Finch contributed the basic mathematical methods, a new point of viev, and a great deal of ecc=cn sence in formulating all of the progrs:s. Masrs. C. R. Ricks and M. Ccvington were respons-ible for the practical aspects of sampling operations personnel perfor-mance.
Mr. J. W. Dykes was a major contributor in the discussions which led to the concept of application of quality control techniques i
to evaluation and control of indus trial work groups.
The greatest credit, however, belongs to those supervisors who
' vere villing to devote their time and suffer the adversity of train-ing outsiders in their organizational problems, sampling info mation, selecting suitable indices and risking demoralization of their people.
In an endeavor of this type, these people accept the risks of failure, do most of the wcrk, and ultimately are much ttre responsible for success or failure of the programs than those vho assist in program fomulation.
These people include Mr. T. H. Stickley, 3.TR/ETR Safety Supervisor, Mr. E. H. Smith, ETR Operations Manager, Mr. F. R. Keller, ATR Operations Manager, and Mr. L. W. Scarbrough, Manager, Maintenance & Experinental Services Branch.
References (1) Werner, SC Ed, The Thyroid 2nd Edition, Page 137, Harper & Row Publishers, New York, Dev York (1962).
0k 0
e 4
6